Validation of the British Society of Thoracic Imaging guidelines for COVID-19 chest radiograph reporting

Area of Science:

• Radiology
• Infectious Disease Imaging
• Thoracic Imaging

Background:

• The COVID-19 pandemic necessitated rapid diagnostic tools.
• Chest radiography is a key imaging modality for COVID-19 assessment.
• Standardized reporting guidelines are crucial for consistent interpretation.

Purpose of the Study:

• To evaluate the reproducibility and diagnostic performance of British Society of Thoracic Imaging guidelines for COVID-19 chest radiograph categorization.
• To assess interobserver agreement among radiologists using these guidelines.

Main Methods:

• Seven consultant radiologists reviewed chest radiographs from 50 COVID-19 patients and 50 controls.
• Radiographs were categorized according to the British Society of Thoracic Imaging guidelines.
• Interobserver agreement was statistically analyzed using Fleiss' kappa and McNemar's tests.

Main Results:

• Excellent specificity (100%) and moderate sensitivity (44%) were observed for the "Classic/Probable COVID-19" category.
• Substantial interobserver agreement (k=0.61) was found for "Classic/Probable COVID-19".
• Fair agreement was noted for "Indeterminate" (k=0.23) and "Non-COVID-19" (k=0.37) categories, with low sensitivity for COVID-19.

Conclusions:

• The proposed guidelines demonstrate high specificity but require refinement for improved sensitivity and interobserver agreement.
• Amalgamating "Indeterminate" and "Non-COVID-19" categories into a "not classic of COVID-19" classification could enhance agreement and reduce misclassification.
• This revised approach would better capture patients with COVID-19 presenting with atypical radiographic findings.